Overclock.net banner
1 - 20 of 48 Posts

·
Premium Member
Joined
·
6,422 Posts
Discussion Starter · #1 · (Edited)
I'm working on a stable RAM overclock and wanted your thoughts. I've always run my Corsair Dominator Platinum kit at it's rated 3200mhz 16-18-18-36 settings. On a whim and for the first time ever, I started some overclock testing yesterday. Currently running (tentatively) stable at 4000mhz overclock with timings tightened to 15-15-15-36. I sent them a bit of juice though and not sure if it's excessive. I think the RAM is defaulted to 1.35v but I cranked it to 1.5v as I've read that some folks run this voltage okay, with my intention of dialing it down once I get a max OC at 1.5v; but I'm clueless when it comes to RAM and voltages. HWinfo64 shows max DIMM temps to be 44C when running the Aida64 memory stress test but not sure if that's hot or not.

Any thoughts, tips? It's TestMem5 Stable, AIDA64 memory stress test stable, Passmark stable but Prime95 fails a worker or two on the Large FFT test ... but I'm not to concerned about Prime95 ... OCCT passes fine too ... CPU is an 8086K running at 5.2ghz / 1.32v ..

Thanks...

~s1rrah
 

·
Registered
Joined
·
657 Posts
Do you know what company and die your Corsair RAM uses? Very important to know as different companies have different limits for voltage.
Corsair tends to use Micron, but not always. For example, in my 18-22-22-42 4,000 MHz kit from them, I've had it running at a constant 1.63V for around a year now.
Probably the tightest possible timings achievable by man on these sticks on my Prime Z390-A. Micron B-die, 64 GB total.
Font Screenshot Parallel Rectangle Software

DDR4 is generally safe up to 1.5V as long as it's not some garbage-tier die. More and more kits are being sold at 1.5V default. It's just that most kits are set to 1.35V since that's all they need to run the sticks.

Best test I've done for stability would be TM5 with either anta777 or 1usmus' config. Honestly, as long as you've tried a good number of tests and have also used the PC normally (gaming, workstation, whatever it is you use it for) and don't experience problems, you're all good. Should do an sfc /scannow every so often to check for any file corruption.

You've got a very good 8086k. Mine can do 5.2 GHz all-core as well, but I gotta stabilize it at like, 1.38V or something. Cache at like, 4.7 GHz? Can't recall.
After getting some instability with the CPU down the line, I just gave up trying and dialed it back down to 5.05 GHz. Difference isn't all that meaningful to me and it is a daily driver.
 
  • Rep+
Reactions: s1rrah

·
Premium Member
Joined
·
6,422 Posts
Discussion Starter · #3 ·
Do you know what company and die your Corsair RAM uses? Very important to know as different companies have different limits for voltage.
Corsair tends to use Micron, but not always. For example, in my 18-22-22-42 4,000 MHz kit from them, I've had it running at a constant 1.63V for around a year now.
Probably the tightest possible timings achievable by man on these sticks on my Prime Z390-A. Micron B-die, 64 GB total.

DDR4 is generally safe up to 1.5V as long as it's not some garbage-tier die. More and more kits are being sold at 1.5V default. It's just that most kits are set to 1.35V since that's all they need to run the sticks.

You've got a very good 8086k. Mine can do 5.2 GHz all-core as well, but I gotta stabilize it at like, 1.38V or something. Cache at like, 4.7 GHz? Can't recall.
After getting some instability with the CPU down the line, I just gave up trying and dialed it back down to 5.05 GHz. Difference isn't all that meaningful to me and it is a daily driver.
Thanks for the comment ... that helps a lot ... especially knowing I have more headroom on the voltages. Been running all night with the sticks bumped up to 3700mhz with the same voltages and timings. Makes me wonder what I could do by running them at 1.5v instead of 1.45v. Fun stuff. The RAM overclock has affected my Cyberpunk 2077 FPS (3840x1600 res/monitor) quite noticeably and in a positive way. Think I've gained bout ~5-7 FPS just by bumping the RAM up to 3700mhz from the stock 3200mhz.

And yeah, this 8086K is something special. I specifically sought out a top 1% binned chip, sourced from iUNLOCK who is quite active over at Notebook Review Forums and does fantastic work binning and customizing desktop CPUs ... he has quite the inventory of top tier chips. This 8086K will actually do 5.2, 5.3 and I've even run 5.4ghz successfully ... all at ridiculously low voltages and temps. It's delidded and LM relidded with a machine lapped, custom all copper IHS. I'd love to see what it could do under LN2 cooling.

Gadget Office equipment Technology Hood Font


I'm currently sitting on a NIB Maximus XIII Hero WiFi / 11700K combo thats on the shelf in front of me and can't decide if I want to "upgrade/sidegrade" to that setup or just stay with this insane 8086K / Maximus X Code setup. A real conundrum as I primarily game on this machine and I don't see any worthwhile FPS gains by moving to the 11th gen setup other than Gen 4 m.2 drives and significantly better multi core performance when rendering, etc.

Check this out ... 8086K @ 5.41 ghz all core overclock with a mere 1.42v. LOL ... it's crazy ... and check the CPU temps throughout, rarely hitting mid 50C range (Corsair H115i XT Pro cooler). This is the Horizon Zero Dawn benchmark at Ultra settings (or High, can't remember). I like it cause it really pushes the CPU compared to some other games. I ran this back when I had the trusty 1080 ti:

8086K @ 5.4ghz / ~1.42v - 1440p Horizon Zero Dawn Bench run


...

Thanks again for the comment .. I'm stoked to try 1.5v on the RAM and see if I can get the mhz/timings better. BTW: in that particular balancing act, MHZ vs. Timings? Which do you think is most important for overall performance, especially in gaming?

Best,

~s1rrah
 

Attachments

·
Premium Member
Joined
·
6,422 Posts
Discussion Starter · #4 ·
3900mhz / 16-16-16-36 / 1.5v ... nipping at the heels of 4000mhz LOL ...
...
Computer Font Screenshot Parallel Software


...

Going to find max MHZ and then work on timings...

Had no idea this kit could do 4000mhz ...

Fun.

~s1rrah
 

·
Premium Member
Joined
·
6,422 Posts
Discussion Starter · #5 · (Edited)
STOCK: 3200mhz / 16-18-18-36 / 1.35v
OVERCLOCK: 4000mhz / 16-16-16-36 / 1.5v

...

Font Screenshot Software Technology Electronic device
 

·
Premium Member
Joined
·
6,422 Posts
Discussion Starter · #6 ·
Passes TestMem5 just fine at 4000mhz ...

Rectangle Product Font Screenshot Parallel
 

·
Registered
Joined
·
657 Posts
My 8086k (which I won in a contest) is delidded with liquid metal as well, although I just reused the same IHS. I did some research before and found that copper/lapping didn't really give any significant gains (margin of error), and I was already capping off at around the 70~80C range with a 240mm radiator AIO on 3DMark's Time Spy Extreme, so I was content. (Most Intel CPUs are perfectly fine up to 80C and most only start experiencing heat-related instability at 85C+.)

The cache clock is very important for RAM overclocking, so try to get it to at least 4.7 GHz if it isn't already. It will boost your memory performance significantly. Do note that the cache clock shares voltage with the core clock and actually demands a lot more of it, so you might find yourself failing to run higher core clocks as you raise the cache clock further and further. You will have to do some experimenting to see where you want to stop. For me, I couldn't get the cache past 4.9 GHz for benchmarking purposes without it experiencing problems.

I'm not hellbent on upgrading my motherboard and CPU for the time being since I'm content with what I have already (and my life would be better spent elsewhere, after all).

I wouldn't recommend pushing above 1.4V for the 8086k since it's part of the older generation of Intel CPUs where degradation is noticable beyond that point. Unless you plan to swap it out after like 2-3 years, I would just stick to like, 5.1-5.2 GHz and call it a day. Up to you, really.

Tightened timings tends to result in more performance than higher frequency. However, it depends entirely on your use-case. For gaming, yeah, I'd suggest tighter timings, but for workstation purposes, bandwidth (higher frequency) can be more important. It's also easier to tighten timings than it is to raise frequency. Try fiddling around with this (lower nanoseconds is better): https://notkyon.moe/ram-latency.htm

I'm not sure if a 15 minute TM5 test is adequate; I personally use 1usmus' config, but I know of many here who use anta777. Most of the time the tests are like, at least an hour long, so you might have to find another anta777 config that does run for that long. Also, if you're a gamer, I highly suggest running and idling in a game that ramps up your GPU to 100% load for at least 2-3 hours (more is better, could even overnight if you want) just to make sure things are stable, because you never know.

Once you're done tightening the primaries, you can delve deeper with the secondaries and tertiaries. Those involve a lot more research and can confuse you a lot of the time, but it's well worth it as they can provide a ton of performance above and beyond just the primaries. IMO though, out of personal experience, leave tRFC and tREFI for last, since they tend to be temperature-bound. The rest of the timings aren't nearly as sensitive.
 
  • Rep+
Reactions: s1rrah

·
Premium Member
Joined
·
6,422 Posts
Discussion Starter · #8 ·
My 8086k (which I won in a contest) is delidded with liquid metal as well, although I just reused the same IHS. I did some research before and found that copper/lapping didn't really give any significant gains (margin of error), and I was already capping off at around the 70~80C range with a 240mm radiator AIO on 3DMark's Time Spy Extreme, so I was content. (Most Intel CPUs are perfectly fine up to 80C and most only start experiencing heat-related instability at 85C+.)

The cache clock is very important for RAM overclocking, so try to get it to at least 4.7 GHz if it isn't already. It will boost your memory performance significantly. Do note that the cache clock shares voltage with the core clock and actually demands a lot more of it, so you might find yourself failing to run higher core clocks as you raise the cache clock further and further. You will have to do some experimenting to see where you want to stop. For me, I couldn't get the cache past 4.9 GHz for benchmarking purposes without it experiencing problems.

I'm not hellbent on upgrading my motherboard and CPU for the time being since I'm content with what I have already (and my life would be better spent elsewhere, after all).

I wouldn't recommend pushing above 1.4V for the 8086k since it's part of the older generation of Intel CPUs where degradation is noticable beyond that point. Unless you plan to swap it out after like 2-3 years, I would just stick to like, 5.1-5.2 GHz and call it a day. Up to you, really.

Tightened timings tends to result in more performance than higher frequency. However, it depends entirely on your use-case. For gaming, yeah, I'd suggest tighter timings, but for workstation purposes, bandwidth (higher frequency) can be more important. It's also easier to tighten timings than it is to raise frequency. Try fiddling around with this (lower nanoseconds is better): https://notkyon.moe/ram-latency.htm

I'm not sure if a 15 minute TM5 test is adequate; I personally use 1usmus' config, but I know of many here who use anta777. Most of the time the tests are like, at least an hour long, so you might have to find another anta777 config that does run for that long. Also, if you're a gamer, I highly suggest running and idling in a game that ramps up your GPU to 100% load for at least 2-3 hours (more is better, could even overnight if you want) just to make sure things are stable, because you never know.

Once you're done tightening the primaries, you can delve deeper with the secondaries and tertiaries. Those involve a lot more research and can confuse you a lot of the time, but it's well worth it as they can provide a ton of performance above and beyond just the primaries. IMO though, out of personal experience, leave tRFC and tREFI for last, since they tend to be temperature-bound. The rest of the timings aren't nearly as sensitive.
I just learned. Thanks. Excellent information. Total mem clocking noob.

~s1rrah
 

·
Premium Member
Joined
·
6,422 Posts
Discussion Starter · #9 ·
I'm not hellbent on upgrading my motherboard and CPU for the time being since I'm content with what I have already (and my life would be better spent elsewhere, after all).
OH yeah ... funny you should say that and I'm in the same boat; I just last week bought a Z590 Asus Maximus Hero / 11700K bundle but am going to return it and keep this 8086K for another year or so as, I'm 90% gaming social browsing and what not and don't do a whole lot of rendering/etc and so can't justify the expense as I pretty much run neck and neck with the 10700K/11700k/11900K folk when it comes to gaming since I play at close to 4K pixel count (3840x1600) ... the gap is much more noticeable at 1080p and slighty noticeable at 1440p but at 1600p/2160p ... I'm generally the same as the 10th/11th gen folk and occasionally 4 to 6 FPS behind on a few titles ... and plus it's just a super fun chip to tinker with ...
 

·
Registered
Joined
·
657 Posts
OH yeah ... funny you should say that and I'm in the same boat; I just last week bought a Z590 Asus Maximus Hero / 11700K bundle but am going to return it and keep this 8086K for another year or so as, I'm 90% gaming social browsing and what not and don't do a whole lot of rendering/etc and so can't justify the expense as I pretty much run neck and neck with the 10700K/11700k/11900K folk when it comes to gaming since I play at close to 4K pixel count (3840x1600) ... the gap is much more noticeable at 1080p and slighty noticeable at 1440p but at 1600p/2160p ... I'm generally the same as the 10th/11th gen folk and occasionally 4 to 6 FPS behind on a few titles ... and plus it's just a super fun chip to tinker with ...
I honestly cannot fault you for wanting to stick with the 8086k; ever since the 9000 series and newer, Intel's been forced to push chips to their absolute limits in order to squeeze out more performance, and they were finally forced to use solder instead for their CPU dies in order to keep temps lower. Even then, many would rant about how their newer CPUs were running extremely hot. And yes, the performance increases are not massive jumps, especially for gaming.

I'd say that the performance only starts to become more noticeable when you are using the PC for non-gaming related tasks, as workstation programs use up more system resources than video games do. Most games are still stuck at using only a few cores/threads at most, meaning people with newer/higher-tier CPUs have a lot of unused cores just sitting around purely working on background tasks. This is why Intel focuses their boost clocks on only a single core or two and then steps the rest down to improve efficiency overall.

For the longest time, budget-savvy gamers would get a decent CPU and then pour the rest of their money into refreshing the GPU every generation or two. That's the most cost-effective way to stay relevant without having to constantly renew the PC. Gaming performance is overall still 80% tied to the GPU, so there's no reason to go out of your way to chase after numbers just for bragging rights. CPU/RAM bottlenecks are uncommon unless you are using something that's insanely old and notice that the GPU can never be fully utilized because the CPU can't handle that many frames. But it's possible to get around that issue by playing on higher resolutions or graphics settings, as it puts more work on the GPU. I know of many who are still happy rocking their 4770k/2600k, with some overclocking.

Keep me posted on how your memory overclocking endeavour goes. I'm pretty sure you have Micron E-die RAM, which is a great overclocker.
 
  • Rep+
Reactions: s1rrah

·
Premium Member
Joined
·
6,422 Posts
Discussion Starter · #11 ·
Keep me posted on how your memory overclocking endeavour goes. I'm pretty sure you have Micron E-die RAM, which is a great overclocker.
Still tweaking ...

Stock: 3200mhz 16.18.18.36
Overclock: 4000mhz 15.15.15.36


Computer Rectangle Font Screenshot Slope



And Thaiphoon Burner identifies the IC's as Samsung but doesn't specify whether C-Die or B-Die ...

Product Font Screenshot Parallel Rectangle


...

15.15.15.34 is next on the agenda ... think I'm going to stay at 4000mhz and work on timings from here on out...

Thanks again for the knowledge...

~s1rrah
 

·
Registered
Joined
·
657 Posts
If it's Samsung, then it's most likely B-die, as their other dies don't overclock nearly as well. Explains why you could make the jump to 4,000 MHz while keeping low primaries.
For Samsung B-die, I wouldn't suggest pushing beyond 1.55V. You can experiment but I've seen mixed results beyond that point. (Micron B-die can tolerate higher voltages much better, which is why I push 1.63V myself on my kit.)

Still suggest using a different test than your anta777 config, since anything under an hour is nowhere near enough to be certain. The RAM sticks aren't heating up enough to throw heat-related errors. Here's the recommended anta777 config, and here's the 1usmus config. Just make a .cfg file in the /bin folder and load it up. It'll take a while to run, but I highly suggest doing so. Leave it on during a lunch break, shopping trip, or overnight if you want.
 

·
Premium Member
Joined
·
6,422 Posts
Discussion Starter · #13 ·
If it's Samsung, then it's most likely B-die, as their other dies don't overclock nearly as well. Explains why you could make the jump to 4,000 MHz while keeping low primaries.
For Samsung B-die, I wouldn't suggest pushing beyond 1.55V. You can experiment but I've seen mixed results beyond that point. (Micron B-die can tolerate higher voltages much better, which is why I push 1.63V myself on my kit.)

Still suggest using a different test than your anta777 config, since anything under an hour is nowhere near enough to be certain. The RAM sticks aren't heating up enough to throw heat-related errors. Here's the recommended anta777 config, and here's the 1usmus config. Just make a .cfg file in the /bin folder and load it up. It'll take a while to run, but I highly suggest doing so. Leave it on during a lunch break, shopping trip, or overnight if you want.
Been working/gaming all day and solid under those circumstances at 4000mhz / 15.15.15.36. Oddly, it posted and tested fine at 15.15.15.34 but read/write speeds were in the mid 56000's instead of the mid/upper 57000's at 15.15.15.36 so went back to that.

Also, tried 4100mhz at 15/36 and hit the wall. Had to happen. Wouldn't post LOL. I'm happy at 4000mhz 15/36 ...

Going to run your suggested tests tonight when I crash ... hopefully the machine doesn't too LOL. I'm not doing anything mission critical with my rig so the fact that it fails a couple Prime95 workers doesn't bother me at all. If it passes OCCT, Aida64 Stability test, etc.? I'm stoked. Going to download your TestMem 5 suggested configs now ... will let you know how it goes ...

Did Cyberpunk 2077 for a couple hours today and RAM temps maxed at 43.3... curious to see what TestMem 5 generates with the extended tests ...

~s1rrah
 

·
Registered
Joined
·
657 Posts
Been working/gaming all day and solid under those circumstances at 4000mhz / 15.15.15.36. Oddly, it posted and tested fine at 15.15.15.34 but read/write speeds were in the mid 56000's instead of the mid/upper 57000's at 15.15.15.36 so went back to that.

Also, tried 4100mhz at 15/36 and hit the wall. Had to happen. Wouldn't post LOL. I'm happy at 4000mhz 15/36 ...

Going to run your suggested tests tonight when I crash ... hopefully the machine doesn't too LOL. I'm not doing anything mission critical with my rig so the fact that it fails a couple Prime95 workers doesn't bother me at all. If it passes OCCT, Aida64 Stability test, etc.? I'm stoked. Going to download your TestMem 5 suggested configs now ... will let you know how it goes ...

Did Cyberpunk 2077 for a couple hours today and RAM temps maxed at 43.3... curious to see what TestMem 5 generates with the extended tests ...

~s1rrah
You can try CL14 for the hell of it, since it either works or it doesn't. After you're done with the primaries, you can move onto the secondaries and tertiaries.

One thing to keep in mind about AIDA64 is that literally anything running in the background (even the OS itself) can cause it to lag at any point in time, thereby reducing the end scores. Hence, I always rerun the test at least a few times per timing change to make sure I get a consistent average rather than just one number.

Stability is in the eye of the beholder; some people would tolerate getting some errors while others wouldn't tolerate any. The main concern of errors is that they can lead to system instability or OS corruption, which is problematic. Be sure to run sfc /scannow every so often to make sure no system files are getting corrupted due to your overclocking, as it can and will happen.

Video games don't really push RAM all that hard in terms of heat; if you do not have a watercooled GPU, the ambient heat dissipated from it can weigh on your RAM, which is why a gaming test is just as important to perform. I don't know how many times my GPU would ramp up to 80C and then my RAM implodes and corrupts system files (my RAM temps are pushed to their absolute limits with a dedicated cooler fan, so...). I've been forced to keep my GPU under 75C at all times, pretty much.
 

·
Premium Member
Joined
·
6,422 Posts
Discussion Starter · #15 ·
Just ran the suggested Anta777 EXTREME test for 1.5 hours. Generated 3 errors starting around 40 minutes. Temps maxed at 45.5C. Might just write that off and run at what I've got or might try backing the timings back to CAS16 and re test ... interesting stuff. Haven't had so much fun overclocking since getting this unicorn level 8086K ... LOL ... more later ...

~s1rrah
 

·
Registered
Joined
·
657 Posts
Just ran the suggested Anta777 EXTREME test for 1.5 hours. Generated 3 errors starting around 40 minutes. Temps maxed at 45.5C. Might just write that off and run at what I've got or might try backing the timings back to CAS16 and re test ... interesting stuff. Haven't had so much fun overclocking since getting this unicorn level 8086K ... LOL ... more later ...

~s1rrah
Yep, sounds like a heat issue. If the 40+ minute mark is beyond the first cycle, then yeah, that's something to address. Maybe additional cooling of some sort.
Try what I did before where I'd just unlid the case and optionally blast a mini USB fan at it. That'll help lower temps a bit to see if you gain more headroom.
 
  • Rep+
Reactions: s1rrah

·
Premium Member
Joined
·
6,422 Posts
Discussion Starter · #17 ·
Yep, sounds like a heat issue. If the 40+ minute mark is beyond the first cycle, then yeah, that's something to address. Maybe additional cooling of some sort.
Try what I did before where I'd just unlid the case and optionally blast a mini USB fan at it. That'll help lower temps a bit to see if you gain more headroom.
So ran the full Anta777Extreme test you suggested over night and it ran for 2:18 hours with 7 errors in test #15 at the 2:15 hour mark. Max temp on one stick was 44.3C, this being the stick closest to CPU socket and the other stick, furthest from the CPU socket only hit 36.5 stock. Interesting.

I'm going to try some cooling solutions first and if that doesn't work then I'll either back the timings off in tiny increments (16.15.15.36, etc. if that's even allowable) or otherwise simply go back to 16.16.16.36 and test again, there).

I have my Corsair H115i as an intake setup so it's blowing fairly warm air directly at the RAM sticks which are probably about 3.5 to 4 inches directly beneath it. I've got lots of cooling headroom on this 8086K as it rarely gets out of the 50C's, even at 5.3ghz so might turn my fans around as outputs as a first and easy step at keeping the RAM cooler.

Another option, and I'm not sure if this will give me better overall speed, is to go to stock timings of 16.18.18.36 and push for something like 4200mhz. Fun balancing act between timings and sheer mhz. Digging the process ...

More later...oh yeah, here is the the log file data for my 2 hour Anta777Extreme run...

Corsair Dominator Platinum DDR4
STOCK: 3200mhz 16.18.18.36 1.35v
TEST OVERCLOCK: 4000MHZ 15.15.15.36 1.51v

========= TestMem5 Log File =========
Customize: Extreme1 @anta777
Start testing at 20:54, 1Gb x12
Error in test #15 through 2:15.16.
Testing is completed through 2:18.57,
detected 7 error(s).

More later...

~s1rrah
 

·
Registered
Joined
·
657 Posts
I'm pretty sure that tRCD+tRP cannot be less than tCL. Also, they don't really influence frequency as much as tCL does, so they probably wouldn't help you push for 4,200 MHz.
A higher vDIMM would likely be more meaningful in that regard. But yes, experiment, since each kit is different.

One thing I dislike about TM5 is that it doesn't log all errors and the types of them; it only logs specific ones and gives you a vague summary in their log files.
You have to pay attention and read the actual error window in the program itself (without closing it) to get the error codes and exact times they occurred, which may not all be Error 15 or at the end.
I've personally not experienced Error 15 myself based on my records, but usually my first suspicion, if you've already cleared one or two cycles successfully, is that it is almost always heat-related.
This is a useful reference for TM5 error codes (the spreadsheet's for AMD users, but the error codes are still relevant): Here

Again, one of the best ways to isolate the issue of heat is to try to blow a fan on the sticks, as I mentioned. It really does make a difference.
One solution (which I kinda hate because it is noisy and there is no other better alternative besides watercooling) is the Corsair Vengeance fan.
It's large and tall enough that it can work for pretty much any RAM, not necessarily their own. Works great, but again, noisy as all hell beyond like, 80% fan speed. Also ugly.
I also know of some who simply twist-tie mini fans to their RAM. That works too, albeit ugly.
 

·
Premium Member
Joined
·
6,422 Posts
Discussion Starter · #19 ·
I'm pretty sure that tRCD+tRP cannot be less than tCL. Also, they don't really influence frequency as much as tCL does, so they probably wouldn't help you push for 4,200 MHz.
A higher vDIMM would likely be more meaningful in that regard. But yes, experiment, since each kit is different.
I'm going to research and begin tweaking the tertiaries eventually ... just ran the EXTREME config you recommended, this time at 16-16-16-36 instead of the former CAS15 settings ... voltage at 1.5V ... I maxed my case fans and took the side panel off of my 780T case and the hottest stick dropped from 45.5C to 42C ... not sure if it was due to the cooler temps or switching to CAS 16 but it passed without errors:

Rectangle Font Screenshot Software Computer


I'm thinking 4000MHZ at those timings is the sweet spot for this kit so going to leave it at that. I tried to play some Cyberpunk 2077 at CAS15 earlier today and it BSoD'd in about 4 minutes. Switched back to CAS16 and played solid for over an hour.

RE: Fan ... I just ordered the Dominator Airflow Platinum RGB RAM cooler and think it will keep things under control temp wise. I've used one before and they work fairly well ... 5 to 10C drop in most cases.

This kit is a version 4.31 kit ... my backup kit of the same ram, bought 5 years ago and never overclocked is version 4.24. This week I'm going to yank these 4000MHZ stable chips out and put the older kit in to check with Thaiphoon Burner for IC type/manufacturer. Might get lucky and have another Samsung kit... 16GB is running things just fine for now, but would like to have the extra headroom of 32gb ...

We'll see...

~s1rrah
 

·
Registered
Joined
·
657 Posts
I'm going to research and begin tweaking the tertiaries eventually ... just ran the EXTREME config you recommended, this time at 16-16-16-36 instead of the former CAS15 settings ... voltage at 1.5V ... I maxed my case fans and took the side panel off of my 780T case and the hottest stick dropped from 45.5C to 42C ... not sure if it was due to the cooler temps or switching to CAS 16 but it passed without errors:

View attachment 2533820

I'm thinking 4000MHZ at those timings is the sweet spot for this kit so going to leave it at that. I tried to play some Cyberpunk 2077 at CAS15 earlier today and it BSoD'd in about 4 minutes. Switched back to CAS16 and played solid for over an hour.

RE: Fan ... I just ordered the Dominator Airflow Platinum RGB RAM cooler and think it will keep things under control temp wise. I've used one before and they work fairly well ... 5 to 10C drop in most cases.

This kit is a version 4.31 kit ... my backup kit of the same ram, bought 5 years ago and never overclocked is version 4.24. This week I'm going to yank these 4000MHZ stable chips out and put the older kit in to check with Thaiphoon Burner for IC type/manufacturer. Might get lucky and have another Samsung kit... 16GB is running things just fine for now, but would like to have the extra headroom of 32gb ...

We'll see...

~s1rrah
I'm sure your kit could be pushed further; it's just a matter of keeping temps under control and gradually tweaking the timings little by little.
It's tedious, arduous, and time-consuming, since you expected to run TM5 after each timing change to check for stability. That's just how things go.
Note down the exact slot order of your current kit to come back to later, as it can affect the kit's overclock-ability.

For me, since I need 64GB for workstation purposes, I had to settle with the best 4x16GB kit I could find, which was this one (and that's basically the only one marked 4,000 MHz+, lol)
It's rated 18-22-22-42 @ 1.35V. Pretty garbage, I know. But somehow, after a month of two of nonstop tweaking, I managed to pull the primaries down to 15-19-19-36 @ 1.63V and also raise the frequency to 4,174 MHz. For Micron B-die, that's pretty damn good. I couldn't go beyond 4,180 or so without my mobo saying nope (that's the best my mid-tier mobo's IMC could drive).
 
  • Rep+
Reactions: s1rrah
1 - 20 of 48 Posts
Top